<![CDATA[Persons with Social Welfare Problems (PMKS) are social groups that live below the community welfare line and are one of component for determining policies in East Java. The study aim to find out the characteristics of the region in East Java based on the PMKS dataset. The method proposed in this study is clustering with the Self-Organizing Maps algorithm and K-Nearest Neighbors (KNN) missing value imputation. KNN used to overcome the amount of missing value in PMKS dataset. First, missing value is filled using KNN imputation. Furthermore, the clustering done with training in SOM network and the result of cluster is evaluated using Silhouette Coefficient. The best parameters for SOM are learning rate=0.1; neighborhood coefficient=0.2; max epoch=160 and neuron size=2x2. The best parameter for KNN is K=2. K=2 gives an increase in Silhouette Coefficient value of 3.4% compared to clustering without missing value imputation KNN. Using best parameter, the highest Silhouette Coefficient obtained is 0.351 which categorized as weak structure. The shape of the cluster produced is a cluster with a proportion of 1:37. The five attributes with the highest difference between the two clusters were Neglected Elderly, Homeless and Psychotic Homeless, Scavengers, Beggars and Minority Groups.]]>
Copyrights © 2019
